ruby bindings: Fix namespacing for libdnf5 classes/modules

[jackorp]

Closes: #1779

The ruby/CMakeLists.txt also defined -module twice. Once "by hand" appending the raw -module ${MODULENAME} and then it is also added via the set_property for SWIG_MODULE_NAME. Once is enough, so I just deleted appending it to CLI and just adjusted the set_property call. This results in all constants in Ruby now prepended with Libdnf5, therefore preventing top-level constant collision (and allowing me to get on with my project).

By the way since setting SWIG_MODULE_NAME appends -module to swig CLI, it is overriding what is present in the files in the %module directive. Both Perl and Ruby are setting -module which makes specifying the module path via %module relatively useless even if there seems to be intent, judging by the fact that they are guarded in module files via #if defined.

JFTR not sure if there is or isn't possibility or some sane way of flattening the namespaces, Logger::Logger::$LOG_LEVEL is a bit unwieldy.

[jrohel]

As a result, all constants in Ruby are now appended with Libdnf5, which prevents top-level constants from colliding (and I can continue my project).

This seems like the right modification. In other languages, we use the libdnf5 namespace. I'm wondering why we don't have that from the start for ruby as well.

I'm also considering how many ruby API users we have. This is an API breaking change. It's easily fixable on the user side. But it is a change.

[jackorp]

This seems like the right modification. In other languages, we use the libdnf5 namespace. I'm wondering why we don't have that from the start for ruby as well.

Seems since "Initial project template" 5bb316d ruby/CMakeLists.txt was wrong or MODULE_NAME had different value than intended judging by the following being present:

+    set_property(SOURCE ../../${LIBRARY_NAME}/${MODULE_NAME}.i PROPERTY SWIG_MODULE_NAME ${MODULE_NAME})
+    set(CMAKE_SWIG_FLAGS -module ${MODULE_NAME} -ruby)
+    swig_add_library(${TARGET_NAME} LANGUAGE ruby SOURCES ../../${LIBRARY_NAME}/${MODULE_NAME}.i)

For comparison, relevant section of perl5 cmake in that same commit:

+    set_property(SOURCE ../../${LIBRARY_NAME}/${MODULE_NAME}.i PROPERTY SWIG_MODULE_NAME ${LIBRARY_NAME}::${MODULE_NAME})
+    swig_add_library(${TARGET_NAME} LANGUAGE perl SOURCES ../../${LIBRARY_NAME}/${MODULE_NAME}.i)

perl does not have the set() directive and set_property() contains ${LIBRARY_NAME}::${MODULE_NAME} instead of just ${MODULE_NAME}

Not sure if there is more history of this somewhere but 2020 is the most I see for these files in this repository.

I'm also considering how many ruby API users we have. This is an API breaking change. It's easily fixable on the user side. But it is a change.

That I do not know, but I can say confidently if they are using libdnf5, they aren't using the default/bundled logger gem that comes with Ruby :D (unless they are jumping through some serious hoops to make it work).

[jrohel]

Thank you for the PR.

We will have to move libdnf5 classes/modules out of the global namespace. This will cause the API change. That's not good. But doing it later will be even more of a pain for all.
Ruby support is in a "best effort" state. From that perspective, it's best to make the change as soon as possible.
I will merge this PR.